Differentiation of Human Limbal-Derived Induced Pluripotent Stem Cells Into Limbal-Like Epithelium

The purpose was to generate human induced pluripotent stem cells (iPSCs) and direct them to limbal differentiation by maintaining them on natural substrata mimicking the native limbal epithelial stem cell (LESC) niche (feederless denuded human amniotic membrane [HAM] and denuded corneas). Limbal-derived iPSCs had fewer unique methylation changes than fibroblast-derived iPSCs, suggesting retention of epigenetic memory during reprogramming, which may facilitate redifferentiation back to limbal cells. Limbal iPSCs cultured for 2 weeks on HAM developed markedly higher expression of putative LESC markers than did fibroblast iPSCs.

Abstract

Limbal epithelial stem cell (LESC) deficiency (LSCD) leads to corneal abnormalities resulting in compromised vision and blindness. LSCD can be potentially treated by transplantation of appropriate cells, which should be easily expandable and bankable. Induced pluripotent stem cells (iPSCs) are a promising source of transplantable LESCs. The purpose of this study was to generate human iPSCs and direct them to limbal differentiation by maintaining them on natural substrata mimicking the native LESC niche, including feederless denuded human amniotic membrane (HAM) and de-epithelialized corneas. These iPSCs were generated with nonintegrating vectors from human primary limbal epithelial cells. This choice of parent cells was supposed to enhance limbal cell differentiation from iPSCs by partial retention of parental epigenetic signatures in iPSCs. When the gene methylation patterns were compared in iPSCs to parental LESCs using Illumina global methylation arrays, limbal-derived iPSCs had fewer unique methylation changes than fibroblast-derived iPSCs, suggesting retention of epigenetic memory during reprogramming. Limbal iPSCs cultured for 2 weeks on HAM developed markedly higher expression of putative LESC markers ABCG2, ΔNp63α, keratins 14, 15, and 17, N-cadherin, and TrkA than did fibroblast iPSCs. On HAM culture, the methylation profiles of select limbal iPSC genes (including NTRK1, coding for TrkA protein) became closer to the parental cells, but fibroblast iPSCs remained closer to parental fibroblasts. On denuded air-lifted corneas, limbal iPSCs even upregulated differentiated corneal keratins 3 and 12. These data emphasize the importance of the natural niche and limbal tissue of origin in generating iPSCs as a LESC source with translational potential for LSCD treatment.

Introduction

Corneal blindness is the second most frequent debilitating eye disorder, affecting 6–8 million people worldwide [1, 2]. Certain corneal diseases are treatable with drugs or surgery [1, 3–5]. However, others are difficult to treat even with corneal transplantation, such as the blinding condition, limbal epithelial stem cell (LESC) deficiency (LSCD). In the human cornea, LESCs reside at the cornea's periphery, the corneoscleral limbus, and continuously renew the epithelium [6–8]. LESC loss can result from congenital aniridia, Stevens-Johnson syndrome, chemical and thermal burns (up to 18% of all eye injuries), ocular cicatricial pemphigoid, chronic inflammation, microbial infections, sulfur mustard gas poisoning, and extended contact lens wear [9–14]. LSCD results in corneal erosions and vascularization, conjunctival ingrowth (conjunctivalization), and scarring, which lead to loss of corneal transparency and, eventually, loss of vision [9, 10].

LESC transplantation can significantly improve vision [15]. However, keratolimbal auto- or allografts for LSCD have a 3–5-year graft survival rate of only 30%–45% [9, 16–18]. Transplanted LESCs expanded in vitro on human amniotic membrane (HAM) or fibrin glue with or without feeder cells can provide success rates of 76% on average for 1–3 years after grafting [6, 7, 18]. Although promising, this procedure is not standardized [19], allograft survival is low, and LESCs can only be propagated for several passages in vitro [19–21]. Potential contamination from mouse 3T3 feeder cells used for culturing LESCs is a further concern [19]. Only scattered attempts are being made to culture LESCs in xenobiotic-free conditions [22, 23] and to optimize cell expansion.

Given the limitations of current LSCD treatments, the development of novel strategies, using renewable and standardized LESC sources, is needed. This is especially important in bilateral LSCD, because patients can only receive transplanted allogeneic LESC. Our approach to this problem builds on advances in induced pluripotent stem cell (iPSC) technology. Unlike cultured limbal stem cells, iPSCs can self-renew for many passages, by expressing high levels of telomerase, allowing continual propagation, banking, and easy cryopreservation [20, 21, 24, 25]. iPSC derivation and growth under good laboratory practice (GLP) and good manufacturing practice (GMP) conditions is rapidly being developed, making these cells poised to move into the clinic to treat a variety of diseases.

We have successfully generated iPSCs from human primary limbal epithelial cells for redifferentiating these iPSCs back into the limbal corneal epithelium. iPSCs show a partial retention of parent cell epigenetic signatures (gene promoter methylation) [25–28]; it can be more prominent in early passage cells [25, 26] but may well persist in late passage iPSCs [28]. Therefore, we used corneal epithelial LESC-enriched cultures to generate iPSCs for programming them into limbal cells, because this strategy might provide an advantage when differentiating iPSCs to a desired lineage [27]. A novel approach for corneal differentiation of iPSCs was to maintain them on a natural niche represented by de-epithelialized human organ-cultured corneas and denuded HAM, closely resembling limbal basement membrane (BM) in composition [7, 20, 21, 29, 30]. This approach has recently shown promise for mouse cells [31]. Using the iPSC technology and advances in cellular reprogramming, a new source of LESCs could be developed with clinical potential for LCSD treatment.

Materials and Methods

Isolation and Culture of Human LESCs

LESC-enriched cultures were prepared from discarded donor corneoscleral rims provided by Y.S.R. and E.M. within 24 hours after corneal transplantation, under an approved Cedars-Sinai Medical Center institutional review board protocol (Pro00019393). Corneoscleral rims with removed conjunctiva were treated with 2.4 U/ml Dispase II (Roche Applied Science, Indianapolis, IN, https://www.roche-applied-science.com) in keratinocyte serum-free medium supplemented with 10% fetal bovine serum (Thermo Fisher Scientific, Life Technologies, Carlsbad, CA, http://www.thermofisher.com) at 37°C for 2 hours [19, 32]. The limbal epithelium was eased off the rims under a microscope, and the cells were dissociated with 0.25% trypsin for 30 minutes at 37°C. Because the LESCs were later used for iPSC generation, the cells were seeded on a mixture of BM proteins [33]: fibronectin, type IV collagen, and laminin (FCL; BD Biosciences, San Diego, CA, http://www.bdbiosciences.com), at 0.5–1 μg/cm².

Decellularization of HAM

With written informed consent, human placentas were obtained after cesarean section deliveries (approved Cedars-Sinai Medical Center institutional review board protocol Pro00019230). HAM isolation and removal of the amniotic epithelium using 0.25 M NaOH have been previously described [34]. For iPSC culture, denuded HAM was placed in CellCrown inserts (Sigma-Aldrich, St. Louis, MO, http://www.sigmaaldrich.com), with the epithelial side facing up. The cells were seeded on the BM surface of HAM. HAM decellularized with NaOH has been shown to support the growth of various cell types [34].

iPSC Generation and Characterization

iPSCs were derived from LESC cultures using a nonintegrating system to avoid random genomic insertion of proviral DNA. Nonintegrating oriP/Epstein-Barr nuclear antigen-1 (EBNA1)-based episomal plasmid vectors derived from Epstein-Barr virus were used to generate transgene integration-free iPSCs. These cells do not have the “footprint” problems experienced when virally inserted DNA is removed after reprogramming [35, 36]. iPSCs were generated from fibroblast or primary LESC cultures by Amaxa Nucleofector transduction (Lonza, Walkersville, MD, http://www.lonza.com) with three episomally expressed oriP/EBNA-1-based pluripotency plasmids from Addgene (Cambridge, MA, http://addgene.org): plasmid pCXLE-hOCT3/4-shp53 (OCT3/4-shp53), plasmid pCXLE-hUL (L-myc-Lin28), and plasmid pCXLE-hSK (SOX2-KLF4) [37]. Five days after transfection, signaling inhibitors were added to enhance reprogramming efficiency and sustain stem cell self-renewal [38, 39] as follows: PD0325901 extracellular signal-regulated kinase inhibitor (Stemgent, Cambridge, MA, http://stemgent.com), CHIR99021 GSK-3β inhibitor, A83-01 transforming growth factor-β receptor, ALK 4/5/7 inhibitor (Tocris Bioscience, R&D Systems, Minneapolis, MN, http://www.tocris.com), and histone deacetylase inhibitor sodium butyrate (Sigma-Aldrich). Colonies grown on growth factor-reduced Matrigel (BD Biosciences) had typical embryonic stem cell (ESC)-like morphology with well-defined borders and a high nuclear/cytoplasmic ratio. After treatment with collagenase/dispase, colonies were lifted off the dish and cultured with dehydroepiandrosterone, epidermal growth factor (EGF), fibroblast growth factor-2, SMAD inhibitors, and heparin to obtain PAX6+ neuroectodermal spheres [40], which have a propensity to give rise to epithelial cells [41]. For differentiation, iPSCs were seeded on different matrices in mTeSR1 medium (StemCell Technologies, Vancouver, BC, Canada, http://www.stemcell.com), and gradually changed to LESC medium (Epilife with defined growth supplements B-27 and N-2, human keratinocyte growth supplements, 10 ng/ml EGF, and an antibiotic-antimycotic mixture; Thermo Fisher Scientific, Life Technologies).

One selected limbal iPSC line from a male donor (0.005% generation efficiency) was named according to the recent nomenclature [42] as CS01iCNL-n.1 (with 1 the clone number) and is abbreviated here as iPSC-L (for limbal). This line was characterized for pluripotency marker expression by immunocytochemistry and PluriTest (a gene-chip and bioinformatics pluripotency assay, based on 23,000-genes Illumina HT12 arrays, and run at the University of California Los Angeles Neuroscience Genomics Core, http://www.semel.ucla.edu/ungc), karyotype by G-banding (at Cedars-Sinai Cytogenetics Core), expression of transduced pluripotency genes by quantitative real-time reverse transcriptase polymerase chain reaction (QRT-PCR), and teratoma formation after cell injection under the kidney capsule of nude mice (under approved institutional review board protocols Pro00028463 and IACUC 004181) [38, 43–45]. Similarly obtained human iPSCs derived from two normal skin fibroblast cell lines, 83iCTR and 14iCTR (from Coriell Institute for Medical Research, Camden, NJ, http://coriell.org [38, 44]), were used for comparison and are abbreviated here as iPSC-83F and iPSC-14F (for fibroblast), respectively. Cells were used at passages 29–43 for iPSC-L, 43–62 for iPSC-83F, and passage 52 for iPSC-14F.

Quantitative RT-PCR

RNA extraction from cultured iPSC-L and iPSC-83F, reverse transcription, and QRT-PCR were conducted as previously described [46]. Published primers were used for plasmid-transduced and endogenous pluripotency genes (OCT3/4, SOX2, KLF4, LIN28, and L-MYC), EBNA-1, and the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (GAPDH) [37]. Primers for putative LESC and corneal epithelial marker genes, PAX6, keratin 3 (KRT3), KRT12, KRT14, KRT15, KRT17, ΔNTP63, and N-cadherin (CDH2) are listed in supplemental online Table 1.

Immunocytochemistry

iPSC-L and iPSC-83F cultured on FCL were fixed for 10 minutes in 100% methanol at −20°C or with 4% p-formaldehyde at room temperature and permeabilized for 10 minutes in 0.2% Triton X-100 (Sigma-Aldrich) in phosphate-buffered saline (PBS) [38, 47]. They were blocked for 1 hour in PBS with 2% bovine serum albumin and 0.05% Triton X-100 at room temperature. Cells cultured on HAM or denuded human corneas were embedded in Tissue-Tek O.C.T. compound (Sakura Finetek USA, Inc, Torrance, CA, http://sakura-americas.com), and cryostat sections were fixed in 2% formalin in PBS for 5 minutes before immunostaining. Indirect single or double immunofluorescent staining was performed as previously published [34, 46, 47]. Primary antibodies are listed in supplemental online Table 2. In some experiments, the number of cells in various lines positive for a specific marker relative to the total number of cells was counted in five random fields of view, and the data were statistically compared by analysis of variance with Bonferroni’s correction for three groups or by the Student t test for two groups using Prism 5.0d software (GraphPad, Inc., San Diego, CA, http://www.graphpad.com). The results are presented as the mean ± SEM, with p < .05 considered significant.

Methylation Array Hybridization

Nine samples for genome-wide methylation analysis were used, including parental cells, iPSCs, and their cultures on HAM. Five hundred nanograms of genomic DNA from each cell culture sample was treated with sodium bisulfite in duplicate, using the EZ96 DNA methylation kit (Zymo Research, Irvine, CA, http://www.zymoresearch.com), following the manufacturer’s protocol. DNA methylation was assessed per the manufacturer’s instructions using the InfiniumHumanMethylation450 BeadChip (Illumina Inc, San Diego, CA, http://www.illumina.com), which interrogated 482,421 DNA methylation sites and covers 99% of Reference Sequence genes, with an average of 17 CpG sites per gene region. In brief, samples were hybridized overnight, washed, and scanned in Illumina iScan (Illumina). Raw array data are available at Gene Expression Omnibus (accession no. GSE53918).

Array Data Analysis

We used methylumi (available at: http://www.bioconductor.org/packages/devel/bioc/html/methylumi.html) to extract β values from Illumina (Illumina Inc) raw idat files. Measurements for which the fluorescent intensity was not statistically significantly greater than the background signal (p > .05) were removed. Methylation intensity β values are reported as a DNA methylation score from 0 (unmethylated) to 1 (completely methylated) using the distribution of β values for probes across the samples. Probes corresponding to the X and Y chromosomes and those containing a single nucleotide polymorphism or a repetitive element within five base pairs of targeted CpG sites were excluded. Therefore, the analysis was restricted to promoter probes mapping to shores. We specifically focused on CpG sites located in a known transcription start site (TSS) region and defined differentially methylated sites as those that, on average, had a methylation state change between two specific cell samples (|Δ|β ≥ 0.2). For visualization of the methylation pattern, a two-way hierarchical clustering using Euclidean distance and average linkage was used. All computations and statistical analyses were performed using R 2.15.0 (R Project for Statistical Computing, Vienna, Austria, http://www.r-project.org) and Bioconductor 2.12 (Bioconductor, based primarily at the Fred Hutchinson Cancer Research Center, Seattle, WA, http://www.bioconductor.org). The genes with differential methylation of CpG sites were compared across the three cell lines using three-way Venn diagrams generated by Vennerable package 3.0 in R 2.15 (R Project for Statistical Computing).

Results

Characterization of Pluripotent Limbal-Derived iPSC-L Line

The parental primary LESCs formed confluent monolayers and expressed putative LESC markers, ABCG2, ΔNp63α, and keratins (K) K15, K17 (Fig. 1), and K14 (data not shown). After reprogramming, the LESC-derived iPSC line (iPSC-L) was found to express the pluripotency markers SSEA-4, TRA-1-81, TRA-1-60, OCT3/4, SOX2, NANOG (Fig. 2A), and alkaline phosphatase (data not shown). Similar patterns were seen in a fibroblast-derived iPSC line (iPSC-83F; Fig. 2A). A PluriTest gene chip assay [45] confirmed pluripotency of all used iPSC lines (supplemental online Fig. 1). A teratoma assay with iPSC-L implanted under nude mice kidney capsules was performed to confirm pluripotency in vivo. After 8 weeks of proliferation, hematoxylin-eosin histologic examination revealed teratomas with the three embryonic germ layers represented by ectodermal neural epithelia, endodermal glandular epithelia of intestinal character, and mesodermal cartilage (Fig. 2A). iPSC-L had a normal male karyotype (46, XY), as shown by G-banding of 20 metaphase cells (Fig. 2B). The correct generation of iPSC requires silencing of exogenous transgenes. In iPSC-L, the expression of exogenous (plasmid) OCT4, SOX2, LIN28, L-MYC, KLF4, and EBNA-1 genes was detectable only at background levels by QRT-PCR using primers specific for either of the exogenous plasmids. However, as expected, endogenous genes were well expressed (Fig. 2C).

Figure 1.

Primary culture of human limbal epithelial stem cell (LESC)-enriched limbal epithelial cells. Note many cells positive for putative LESC markers, K15, K17, ABCG2, and ΔNp63α (bright). Immunofluorescent staining. Scale bar = 30 μm. Abbreviation: K, keratin.

Figure 2.

Characterization of human limbal iPSC-L line. (A): Pluripotency testing. Upper row: Positive staining for multiple pluripotency markers in iPSC-L cultures. Middle row: Similar marker patterns in a fibroblast-derived iPSC-83F line. Bottom row: Limbal iPSC-L gives rise to teratomas in immunodeficient mice that form ectoderm, endoderm, and mesoderm. Scale bars in upper and middle rows = 100 μm; in bottom row = 60 μm. Nuclei are counterstained with DAPI. (B): Diploid male karyotype of limbal iPSC-L, G-banding. (C): Quantitative reverse transcriptase-polymerase chain reaction of pluripotency genes in iPSC-L (passage 3). Note expression of endogenous genes and only negligible expression of plasmid-derived genes. Bar values were calculated relative to endogenous genes in fibroblast iPSC-83F. Abbreviations: DAPI, 4′,6-diamidino-2-phenylindole; iPSC-83F, iPSCs from normal skin fibroblast cell line 83iCTR; iPSC-L, CS01iCNL-n.1.

Differentiation of iPSC Lines on Limbal Basement Membrane-Like Extracellular Matrix

To differentiate iPSCs toward a limbal fate, we focused on extracellular niches important for stem cell maintenance, limbal differentiation of embryonic and hair follicle stem cells, and iPSC [33, 48–51]. Limbal BM contacting LESCs in vivo is a complex assembly of laminins, type IV collagens, glycoproteins (including fibronectin), and proteoglycans [52–54]. Therefore, we first tried to differentiate iPSCs toward LESC by culturing them on the FCL (fibronectin-type IV collagen-laminin) mixture of BM components that are part of limbal BM.

Immunostaining for putative LESC markers was performed in iPSC-L and iPSC-83F after 15 days in culture on FCL. Similar to cultured LESCs, limbal iPSC-L showed distinct expression of K14 and K15 (Fig. 3A, 3B). However, control fibroblast iPSC-83F cells (Fig. 3A) expressed significantly less K15 and, especially, K14 (Fig. 3B), suggesting that limbal BM-like matrix supported epithelial differentiation better in limbal iPSCs than in fibroblast iPSCs. Limbal iPSC-L also showed stronger and more uniform staining for a LESC marker, ΔNp63α, compared with fibroblast iPSC-83F (Fig. 3A, 3B). Clinically, a percentage of high ΔNp63α expressing cells is used as a prognostic marker in LESC transplantation [6]. Overall, limbal-derived iPSCs on the FCL matrix showed a closer differentiation to LESC than fibroblast-derived iPSCs, as judged by the expression of several LESC markers.

Figure 3.

Cultures of human LESC, iPSC-L, and iPSC-83F on FCL (fibronectin, type IV collagen, and laminin) matrix. (A): Immunostaining for putative LESC markers, K14, K15, and ΔNp63α. Note similar expression of K14 and K15 in LESC and limbal iPSC-L and markedly weaker expression in fibroblast iPSC-83F. Also, more cells are ΔNp63α-bright in iPSC-L than in iPSC-83F 2-week cultures. Nuclei are counterstained with DAPI. Antibodies to K15 and ΔNp63α were used for double staining; DAPI staining corresponding to ΔNp63α is shown on respective K15 panels. Scale bar = 40 μm. (B): Quantitation of positive cells in cultures. Data are expressed as ratios of the number of marker expressing cells (bright staining only for ΔNp63α [6]) to total number of cells in the same field revealed by DAPI nuclear staining. Asterisks denote significance levels: ∗, p < .05; ∗∗, p < .01; ∗∗∗, p < .001; ∗∗∗∗, p < .0001. Abbreviations: DAPI, 4′,6-diamidino-2-phenylindole; iPSC-83F, iPSCs from normal skin fibroblast cell line 83iCTR; iPSC-L, CS01iCNL-n.1; LESC, limbal epithelial stem cell; K, keratin.

Differentiation of iPSCs Toward LESC Lineage on Native Biological Substrata

We next examined whether a more natural LESC niche would better support iPSC differentiation toward the limbal epithelial lineage. Exposed HAM BM largely resembles native limbal BM and has been commonly used as a support for seeding LESC for subsequent human transplantation [34, 55]. Limbal iPSC-L and fibroblast iPSC-83F were seeded on NaOH-denuded HAM and grown for up to 2 weeks. QRT-PCR of 8-day (data not shown) and 15-day iPSC cultures on HAM showed increased expression of several corneal keratin genes (KRT3, KRT14, and KRT17) and some genes encoding putative LESC markers (ΔNTP63, PAX6, and CDH2/N-cadherin) compared with undifferentiated iPSCs (Fig. 4). The expression levels of these genes were consistently higher in limbal iPSC-L than in fibroblast iPSC-83F (Fig. 4). These data were further verified by immunostaining using sections of HAM with 15-day-old iPSC cultures. HAM-grown iPSC-L cultures developed good expression of putative LESC markers, K14, K15, tropomyosin-receptor-kinase A (TrkA), ΔNp63α, N-cadherin (Fig. 5), and K17 (data not shown). In contrast, control fibroblast iPSC-83F showed little expression of TrkA, K14, ΔNp63α (Fig. 5), or K15 (data not shown). Importantly, immunostaining for the conjunctival marker K13 revealed its insignificant expression in iPSC-L (Fig. 5).

Figure 4.

Quantitative reverse transcriptase-polymerase chain reaction of putative limbal epithelial stem cell (LESC) marker genes in induced pluripotent stem cell cultures on fibronectin, type IV collagen, and laminin (FCL) (0 days) and denuded human amniotic membrane (HAM) (15 days). Initial cultures on FCL were transferred to HAM on day 0. HAM cultures have increased expression of all genes but decreased expression of a pluripotency marker gene OCT4. Note higher expression (log scale) of LESC marker genes in limbal iPSC-L compared with fibroblast iPSC-83F after 15 days of culture on HAM. Abbreviations: CDH2, N-cadherin; d, day; iPSC-83F, iPSCs from normal skin fibroblast cell line 83iCTR; iPSC-L, CS01iCNL-n.1; LESC, limbal epithelial stem cell; KRT3, keratin 3; KRT14, keratin 14; KRT17, keratin 17; mRNA, messenger RNA.

Figure 5.

Immunostaining of putative limbal epithelial stem cell markers in 2-week induced pluripotent stem cell cultures on denuded human amniotic membrane. Note distinct to strong expression of K14, K15, TrkA, N-cadherin, and ΔNp63α in limbal iPSC-L compared with weak expression of K14, TrkA, and ΔNp63α in fibroblast iPSC-83F. Only negligible staining is observed in iPSC-L for a conjunctival keratin K13. Scale bar = 30 μm. Abbreviations: iPSC-83F, iPSCs from normal skin fibroblast cell line 83iCTR; iPSC-L, CS01iCNL-n.1; LESC, limbal epithelial stem cell; K, keratin.

In order to verify iPSC differentiation using a natural LESC niche, iPSC-L were seeded on the surface of organ-cultured human corneas [56, 57], with NaOH-removed corneal and conjunctival epithelium. To facilitate cell differentiation, the corneas were air-lifted [58] by keeping the culture medium at the limbus level. After 2 weeks of culture, the iPSC-L marker expression profiles were examined. The iPSC-L had proliferated and spread on the corneal surface (Fig. 6). They also expressed putative LESC markers, ABCG2, K14, TrkA (Fig. 6), and K17 (data not shown). Importantly, they also expressed markers of differentiated corneal epithelium, K3 and K12 (Fig. 6). These results suggest that (a) growing iPSC on their preferred niche (HAM and, especially, corneal surface) helps drive their corneolimbal differentiation, and (b) the chosen niche enhances epithelial differentiation from limbal-derived iPSCs better than from fibroblast-derived iPSCs.

Figure 6.

Differentiation of limbal-derived CS01iCNL-n.1 (iPSC-L) on denuded human cornea. Cells were cultured for 2 weeks on the epithelial side of NaOH-denuded human cornea. Upper left panel: Phase contrast view of iPSC-L partially occupying the corneal surface. Note strong expression of putative limbal epithelial stem cell markers, ABCG2, K14, TrkA, and of differentiated corneal epithelial markers K3 and K12 in the induced pluripotent stem cells (arrows) adhered to the human corneal surface. Nuclei are counterstained with DAPI. Scale bar = 40 μm. Abbreviations: DAPI, 4′,6-diamidino-2-phenylindole; K, keratin; S, stroma.

Global Methylation Profiles of iPSC Lines

That limbal iPSCs more readily expressed limbal epithelial markers than did fibroblast iPSCs when cultured on limbal BM-like matrix (Figs. 4, 5) could have resulted from the known partial retention of parent tissue epigenetic signatures by iPSCs [26–28, 59, 60], which could facilitate LESC generation from limbal iPSCs. Therefore, we next examined the epigenetic signatures, as exemplified by DNA methylation patterns, of iPSC lines compared with parental cells.

Global promoter methylation analysis revealed modest changes in three independent iPSC lines (one limbal-derived and two fibroblast-derived) compared with the parental cells. By supervised analysis, line-specific hypomethylation and hypermethylation changes vs. parental cells involved approximately 3,000 genes in each iPSC, similar to previously reported numbers [61]. A significant overlap of genes with the methylation changes was seen among the three iPSC lines, iPSC-L, iPSC-83F, and iPSC-14F. These changes might reflect general deviation of iPSCs from tissue cells during acquisition of pluripotency. Therefore, we focused on methylation changes unique to each iPSC line compared with primary parental cells (limbal epithelial cells, and fibroblasts 83i and 14i). As shown in supplemental online Table 3 and Venn diagrams (supplemental online Fig. 2), limbal iPSC-L had only 64 unique genes with changed methylation compared with the parental limbal cells. This number increased to 101 genes (up 57.8%) when limbal iPSC-L was compared with 83i primary fibroblasts and to 105 genes (up 64.1%) when compared with 14i fibroblasts. These changes were observed for both hypermethylated and hypomethylated genes. In fibroblast-derived iPSC-83F (supplemental online Table 3; supplemental online Fig. 3), 76 genes showed unique methylation differences compared with parental 83i cells. This number increased to 101 genes (up 32.9%) when iPSC-83F was compared with 14i fibroblasts and to 79 genes (up 3.9%) when compared with limbal cells. These data show that the unique methylation patterns of limbal-derived iPSCs were more similar to the parental cells than to skin fibroblasts. Similarity to parental cells was also observed in fibroblast iPSC-83F. However, iPSC-14F (supplemental online Table 3; supplemental online Fig. 4) showed 652 uniquely changed genes compared with parental 14i fibroblasts, which was not significantly different from the uniquely changed genes vs. 83i fibroblasts (634; down 2.8%), or even vs. limbal cells (600; down 8.7%). These data illustrate significant heterogeneity of iPSCs in terms of methylation patterns (see also [61]) and suggest that not all iPSC lines will retain certain profiles of parental cells.

Methylation Profiles of iPSC Lines Cultured on HAM

Hierarchical cluster analysis was performed to assess whether differentiation on HAM changed the methylation patterns of some iPSC genes toward the parental cells. Promoter region probes were significantly enriched in probes mapping to shores of CpG islands compared with the total number of probes in the 450K platform (p < .0001; Fisher’s exact test). Therefore, the analysis was restricted to probes mapping to shores. A subset of data was created from probes mapping to promoter region 1,500 base pair (bp) upstream and downstream of TSS.

Probes that showed a β value of 0.15 or more in at least one sample from data corresponding to primary limbal cells, fibroblast 83i, iPSC-83F, and iPSC-83F on HAM were used for further analysis. Probes corresponding to shores within mapped regions yielded 2,309 genes. A set of 504 genes was further selected using stringent criteria of defining differential methylation (hyper- and hypomethylation) in a three two-group (|Δ|β ≥ 0.2) comparison. This threshold corresponds to the recommended difference between groups of samples analyzed with the Illumina methylation Infinium technology (Illumina Inc) that can be detected with 99% confidence and presents a stringent estimate of Δβ detection sensitivity across the range of β-values [62, 63]. A two-way hierarchical analysis showed a tight clustering of iPSC-83F and their derivatives cultured on HAM (Pearson correlation coefficient = 0.932) that was distinct from the methylation profile of primary limbal cells (Fig. 7A). In particular, one cluster of 243 genes showed a very distinct pattern of methylation common to fibroblast 83i, iPSC-83F, and iPSC-83F on HAM but distinctly different from primary limbal cells (Fig. 7A; supplemental online Fig. 5A).

Figure 7.

Hierarchical cluster analysis of methylation profiles of select genes in various induced pluripotent stem cells and parental cells. (A): Comparison of fibroblast 83i-derived cells with primary limbal cells. Left: Heat map of 504 methylated genes (0, unmethylated; 1, methylated) selected using stringent criteria of defining hyper- and hypomethylation in three two-group (|Δ|β ≥ 0.2) comparisons. Fibroblast 83i, iPSC-83F, and iPSC-83F on HAM show similarity to each other but clear differences from primary limbal cells. Right: Analysis of a 243-gene cluster (marked with asterisk on left panel) showing an excellent correlation of methylation patterns in fibroblast 83i, iPSC-83F, and iPSC-83F on HAM. HAM culture did not drive methylation profiles of these genes closer to limbal cells, and they remained close to parental fibroblasts. (B): Comparison of limbal-derived cells with fibroblast-derived iPSC-83F on HAM. Left: Heat map of a subset of 3,376 methylated genes, which were differentially methylated, identified by analyzing probes of genes that showed differential methylation in at least one- to two-group comparisons. Right: A subcluster of 129 genes (marked with asterisk on left panel) with their methylation profiles in iPSC-L on HAM closely correlating with primary limbal cells and primary limbal cells on HAM but different from original iPSC-L and even more different from iPSC-83F on HAM. HAM culture drove methylation profiles of these genes closer to parental limbal cells and away from original iPSC-L. Abbreviations: HAM, human amniotic membrane; iPSC-83F, iPSCs from normal skin fibroblast cell line 83iCTR; iPSC-L, CS01iCNL-n.1.

Similarly, in primary limbal cells, the same cells on HAM, iPSC-L on HAM, and iPSC-83F on HAM, a subset of 3,376 genes was identified that were differentially methylated. Here, probes of genes that showed differential methylation compared with at least one to two groups were accounted for. Figure 7B shows a hierarchical clustering of 3,376 genes displaying differential methylation patterns between the primary limbal cells and primary limbal cells on HAM, iPSC-L and iPSC-L on HAM, and iPSC-83F on HAM. A subcluster of 129 genes had their methylation profiles in iPSC-L on HAM closely correlated with primary limbal cells and primary limbal cells on HAM but was different from iPSC-L and even more different from iPSC-83F on HAM (Fig. 7B; supplemental online Fig. 5B). Primary limbal cells acquired very few gene methylation changes on culture on HAM (Pearson correlation coefficient = 0.943). Cluster analysis thus identified a distinct subset of genes in iPSC-83F that had similar methylation patterns to primary 83i fibroblasts and did not become closer to limbal cells on iPSC-83F culture on HAM. In contrast, another subset of genes in original iPSC-L differed in the methylation patterns from the primary limbal cells but became very similar to these parental cells on iPSC-L culture on HAM. The expression of these genes could drive apparent differentiation of iPSC-L into limbal lineage on HAM, judged by LESC marker expression, contrary to fibroblast iPSC-83F.

Methylation Profiles of LESC-Associated Genes in iPSCs and Parental Cells

A number of putative LESC markers have been identified with preferential location in the LESC compartment. We compared the methylation profiles of 28 LESC-associated genes [64–80] (supplemental online Table 4) and two differentiated corneal epithelial genes coding for K3 and K12 (KRT3 and KRT12) in limbal and fibroblast iPSCs. In this analysis, we identified those genes that did not differ in methylation (both hyper- and hypomethylation) in select iPSCs compared with their parental cells and then determined their methylation changes in unrelated iPSCs. Compared with iPSC-L and parental limbal cells, iPSC-83F had the CDH3 gene hypermethylated and LGR5 gene hypomethylated. iPSC-14F had the LGR5 gene hypomethylated. Compared with iPSC-83F and 83i cells, iPSC-L had the KRT19 gene hypermethylated. iPSC-14F also had the KRT19 gene hypermethylated and LAMC3 gene hypomethylated. Compared with iPSC-14F and 14i cells, iPSC-L had 4 genes hypermethylated. iPSC-83F had 5 genes hypermethylated and the EREG gene hypomethylated (supplemental online Table 5).

The methylation patterns of LESC-associated genes were also examined after cell culture for 2 weeks on denuded HAM. Compared with primary LESC, iPSC-L on HAM differed from the same original cells cultured on FCL only by change of the NTRK1 gene from hypermethylated to hypomethylated (presumably, activated) state (supplemental online Table 6). Interestingly, the NTRK1 gene became hypermethylated in iPSC-83F on HAM (not the same as in iPSC-83F on FCL but the same as in the parental fibroblasts), unlike iPSC-L on HAM, probably accounting for the reduced expression of its product, TrkA, compared with iPSC-L on HAM (Fig. 5). Other differences between fibroblast-derived and limbal-derived iPSCs on HAM relative to primary LESC included hypomethylation of the LGR5 and SOD2 genes and a lack of hypomethylation of the SLC2A1 gene in iPSC-83F (supplemental online Table 6).

Compared with primary limbal cells, parental 83i fibroblasts and iPSC-83F on FCL (original) and HAM had changes in a number of genes. The difference of iPSC-83F on HAM from the original iPSC-83F was in 2 hypermethylated genes and 4 hypomethylated genes. In contrast, its difference from 83i was in 5 hypermethylated and 8 hypomethylated genes (supplemental online Table 7). iPSC-83F on HAM was still closer to the original iPSC-83F than to the parental 83i fibroblasts in methylated LESC-associated genes in contrast to the limbal cells. Thus, iPSC-L on HAM became somewhat closer to parental limbal cells, unlike the fibroblast-derived iPSC-83F on HAM. This result was in agreement with immunostaining of iPSC cultures on HAM for LESC-associated proteins and gene cluster analysis.

Discussion

Despite some success in treating LSCD with LESC biopsies or cultures [6, 81, 82], the need still exists for renewable and standardized transplantable LESC sources, especially when a patient’s own LESCs are unavailable for autologous grafting. Differentiated iPSCs could be a promising source of LESCs, similar to lens epithelial and retinal pigment epithelial (RPE) cells and photoreceptors [83, 84]. However, the differentiation of stem cells into corneal epithelium has proved to be difficult. Notably, ESCs seeded on corneas with partially removed epithelium have failed to stratify [50, 85]. Fibroblast-derived iPSCs grown on denuded HAM differentiated mostly into neural precursors and RPE cells [86]. Fibroblast iPSCs seeded on mouse cornea and differentiated with bone morphogenetic protein produced stratified epithelium expressing p63 and K15 but also expressed epidermal markers [31]. The only promising protocol for differentiating mouse ESCs toward the limbal lineage used conditioned media from limbal fibroblasts, which resulted in the expression of corneal keratin K12 and the formation of stratified epithelium on a pig cornea [87]. In another study, limbal-derived iPSCs were differentiated into limbal-like epithelium after several weeks in culture using the “stromal cell-derived inducing activity” method [61]. Although limbal-derived iPSCs did express corneal keratins, including K3/K12, some corneal marker expression was also noted in the fibroblast-derived iPSCs with this method. In addition, limbal- and fibroblast-derived iPSCs showed no significant difference in the methylation patterns of the LESC-associated genes PAX6, TP63, and K14 and corneal epithelial genes K12 and K3 [61]. Overall, lens, RPE, and photoreceptor cells have been successfully produced from ESCs or iPSCs; however, optimized protocols for making corneal epithelium from stem cells are missing. This could be because of the use of noncorneal stem cell source or differentiation on a noncorneal niche (feeder cells, coated plates).

We describe a different approach, differentiating LESC-derived iPSCs without any feeder cells but using biological supports that are similar (denuded HAM) or identical (denuded organ-cultured corneas) to their natural niche. The importance of niche factors, including underlying BM and stromal cells, has been emphasized [8, 88]. The limbal fibroblast addition to cultured LESC was shown to increase their growth potential [89]. Isolating LESCs from corneas with their stromal niche cells resulted in increased clonogenicity, even without 3T3 feeder cells [90]. As an alternative to the corneal surface, HAM is widely used for LESC propagation. It resembles the limbal BM, contains important growth factors, and is nonimmunogenic [7, 29, 30, 54, 55, 91]. HAM is considered the best substratum for LESC expansion [7], also allowing secure placement onto the patient’s cornea.

To test the niche hypothesis, we used reprogrammed diploid limbal iPSCs that had passed stringent pluripotency tests. Cloned cells formed large outgrowths when seeded on NaOH-denuded HAM. After 2 weeks, limbal-derived iPSCs on HAM expressed putative LESC markers, ΔNp63α, N-cadherin, TrkA, and keratins K14, K15, and K17 but did not express the conjunctival marker K13 (Fig. 5). On de-epithelialized human corneas, even more advanced differentiation of limbal iPSCs was obtained, evidenced by the expression of the mature corneal epithelial markers K3 and K12 (Fig. 6). However, expression of LESC-associated markers was not pronounced in fibroblast-derived iPSCs on HAM. These data attest to the advantage of using natural support/niche rather than animal feeder cells [61] as a potent factor facilitating redifferentiation of limbal iPSCs into LESC-like cells. Natural substrata, including denuded HAM and corneal surface, are apparently better than BM protein mixture or Matrigel (BD Biosciences) because their composition and the presence of stromal cells resemble the LESC niche more closely [8, 33, 48, 55, 89, 90]. HAM is also Food and Drug Administration-approved, facilitating translation of HAM-differentiated iPSC-derived LESCs to the clinic. The appearance of differentiated corneal keratins when iPSCs were kept air-lifted on the corneal surface provides a good prospect that even a stratified epithelium suitable for transplantation could be obtained from iPSCs using our approach.

Optimized LESC differentiation from limbal-derived iPSC might depend on both the proper niche and the specific soluble factors regulating the LESC state and function (e.g., LESC-expressed stromal cell-derived factor 1/C-X-C motif chemokine 12, and its limbal fibroblast receptor C-X-C chemokine receptor 4 [88], Wnt/β-catenin system and transcription factor 4 [32, 92], keratinocyte and hepatocyte growth factors [93, 94], transforming growth factor-β [95], and some microRNAs [96]).

An additional factor for successful differentiation of LESC-like cells from iPSCs might be using the same tissue of origin [97]. Derivation of iPSCs from cultured LESCs rather than from other cells types could facilitate their redifferentiation back to LESCs owing to retention of methylation-related epigenetic signatures [25–28, 61]. This fact is highlighted by the higher expression of LESC markers by differentiating iPSCs reprogrammed from limbal epithelium rather than from fibroblasts. Soluble factor-mediated differentiation of limbal and fibroblast iPSCs produced only very modest quantitative differences in epigenetic DNA methylation profiles between them, with no difference in methylation for a few LESC-associated genes [61]. Using limbal-like niches as iPSC substrata, we identified changes in the methylation of genes unique to each iPSC line. In limbal iPSC-L and fibroblast iPSC-83F, these changes were less pronounced vs. parental cells compared with the irrelevant primary cells. However, the fibroblast iPSC-14F line deviated from all parental cells. Therefore, iPSC differentiation to the desired lineage could be facilitated by the selection of clones more similar to the parental cells in the methylation profiles. Importantly, residual epigenetic signatures can be retained in advanced passage iPSCs, as published data [28] and our present work have shown. However, early passage iPSCs that might be more similar to parental cells [25, 26] could be easier to redifferentiate back to the parental tissue.

During differentiation on HAM, a subset of genes in limbal-derived iPSCs had their methylation profiles shifted closer to the parental cells compared with the initial iPSCs. At the same time, fibroblast-derived iPSC-83F on HAM had a cluster of genes with methylation profiles not changed toward limbal cells but close to the original iPSCs and parental fibroblasts. HAM, similar to limbal BM, might thus facilitate preferential differentiation to the limbal lineage of the limbal-derived iPSCs in contrast to fibroblast-derived iPSCs. The data suggest that denuded corneas would be similar to HAM in this respect.

A previous report [61] found that limbal- and fibroblast-derived iPSCs were similar in methylation of a small number of LESC-associated genes studied. In the present study, methylation similarities in many of the 28 LESC-associated genes analyzed were also observed between the limbal- and fibroblast-derived iPSCs differentiated on HAM (supplemental online Table 6), confirming the idea that many putative stem cell markers might not be strictly LESC-specific [98]. However, several notable differences were found, concerning NTRK1, SLC2A1, LGR5, and SOD2 genes. In general, limbal-derived iPSC-L and fibroblast-derived iPSC-83F both appeared to be more similar to their parental cells in the methylation profiles of this gene group. On HAM, fibroblast-derived iPSCs remained close to the original iPSCs but not to the limbal cells or limbal-derived iPSCs. Noteworthy, NTRK1 gene coding for the nerve growth factor receptor TrkA was hypermethylated in limbal iPSC-L compared with parental LESCs but became hypomethylated (possibly activated) vs. LESC in these cells after culture on HAM. In contrast, in fibroblast iPSC-83F on HAM, the NTRK1 gene became hypermethylated compared with the original iPSC-83F and closer to the parental 83i cells (supplemental online Table 7). This methylation difference correlated with markedly higher TrkA protein expression in iPSC-L on HAM compared with iPSC-83F on HAM (Fig. 5). Corneal TrkA is preferentially expressed in basal limbal cells [75]. The data suggest that its activation (e.g., by hypomethylation) might be important for LESC maintenance and iPSC differentiation. As such, it could become a useful marker for monitoring limbal epithelial differentiation of iPSC. Conversely, in iPSC-83F on HAM, the hypomethylated status of SLC2A1 gene coding for the glucose transporter 1 (GLUT1) was lost on HAM culture, unlike in iPSC-L. This change in iPSC-83F could be related to differential regulation of GLUT1 in epithelial and fibroblast cells [99], which was also documented for the SOD2 gene product [100], showing hypomethylation on HAM only in iPSC-83F.

The choice of limbal tissue for the generation of iPSCs and their redifferentiation into LESCs with the proper extracellular niche could significantly advance the translation of iPSC-derived LESCs to the treatment of LSCD. This applies, in particular, to bilateral LSCD cases for which autologous transplantation is not feasible, and a reliable banked source of LESC is needed.

Conclusion

Limbal epithelium-derived iPSCs differentiated into limbal-like cells more readily than did fibroblast-derived iPSCs when cultured on natural substrata mimicking native limbal stem cell niche (denuded HAM or corneal surface). This might be due to partial retention of parental tissue epigenetic signatures, including the methylation patterns of some LESC-associated genes (e.g., NTRK1). Such iPSCs could become a new expandable and bankable LESC source for transplantation to patients with LSCD.

Author Contributions

D.S., M.S.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript; L.O., M.A.W., K.N., A.S., L.S.: collection and/or assembly of data, final approval of manuscript; V.A.F.: data analysis and interpretation, manuscript writing, final approval of manuscript; J.T. and V.P.: data analysis and interpretation, final approval of manuscript; E.M.: provision of study material or patients, final approval of manuscript; Y.S.R.: conception and design, provision of study material or patients, final approval of manuscript; C.N.S.: conception and design, data analysis and interpretation, financial support, administrative support, manuscript writing, final approval of manuscript; A.V.L.: conception and design, data analysis and interpretation, manuscript writing, financial support, final approval of manuscript.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.